Time-series classification of Sentinel-1 agricultural data over North Dakota

Abstract

ABSTRACTAccurate measurements of agricultural land cover are important for monitoring global food security, economic stability, and environmental conditions. Since significant portions of global agricultural land are frequently cloud covered, synthetic aperture radar (SAR) has been shown to be a reliable form of gathering crop measurements, even in regions where acquiring clear optical imagery is challenging. In this work, repeat coverage from the C-band Sentinel-1 satellite over a portion of North Dakota is used to classify individual agricultural land-cover types. In this approach the times series forms the basis of a classification algorithm, where individual pixels are compared against a model of average crop backscatter response and classified as the crop with the least difference from the model. Multiple variations on the analysis are run to test the influence of polarization, iterations in model building, number of training fields, and validation input on the classification accuracy. It is shown that both VV and VH polarizations individually and combined are routinely able to produce overall accuracies above 90% when using multiple iterations in model building. These results show the potential for SAR-based agricultural land cover classifications built from comprehensive time series data.